التفاصيل البيبلوغرافية
| العنوان: |
Learning from Biased Data: A Semi-Parametric Approach |
| المؤلفون: |
Bertail, Patrice, Clémençon, Stéphan, Guyonvarch, Yannick, Noiry, Nathan |
| المساهمون: |
Fédération Parisienne de Modélisation Mathématique (FP2M), Centre National de la Recherche Scientifique (CNRS), Modélisation aléatoire de Paris X (MODAL'X), Université Paris Nanterre (UPN)-Centre National de la Recherche Scientifique (CNRS), Signal, Statistique et Apprentissage (S2A), Laboratoire Traitement et Communication de l'Information (LTCI), Institut Mines-Télécom Paris (IMT)-Télécom Paris-Institut Mines-Télécom Paris (IMT)-Télécom Paris, Département Images, Données, Signal (IDS), Télécom ParisTech, Paris-Saclay Applied Economics (UMR PSAE), AgroParisTech-Université Paris-Saclay-Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement (INRAE), International Conference on Machine Learning |
| المصدر: |
Proceedings of Machine Learning Research ; ICML 2021 virtual conference - 38th International Conference on Machine Learning ; https://hal.inrae.fr/hal-04431531 ; ICML 2021 virtual conference - 38th International Conference on Machine Learning, International Conference on Machine Learning, Jul 2021, En ligne (Etats-Unis), United States. pp.803-812 ; https://icml.cc/virtual/2021/index.html |
| بيانات النشر: |
HAL CCSD |
| سنة النشر: |
2021 |
| مصطلحات موضوعية: |
transfer learning, generalized method of moments, reweighted risk minimization, [SHS.ECO]Humanities and Social Sciences/Economics and Finance |
| جغرافية الموضوع: |
En ligne (Etats-Unis), United States |
| الوصف: |
International audience ; We consider risk minimization problems where the (source) distribution P-S of the training obser- vations Z(1),., Z(n) differs from the (target) distribution P-T involved in the risk that one seeks to minimize Under the natural assumption that P-S dominates P-T , i.e. PT << PS, we develop a semiparametric framework in the situation where we do not observe any sample from P-T, but rather have access to some auxiliary information at the target population scale. More precisely, assuming that the Radon-Nikodym derivative dP(T)/dP(S)(z) belongs to a parametric class {g(z, alpha), alpha is an element of A} and that some (generalized) moments of P-T are available to the learner, we propose a two-step learning procedure to perform the risk minimization task. We first select (alpha) over cap so as to match the moment constraints as closely as possible and then reweight each (biased) training observation Z(i) by g(Z(i), (alpha) over cap) in the final Empirical Risk Minimization (ERM) algorithm. We establish a O-P(1/ root n) generalization bound proving that, remarkably, the solution to the weighted ERM problem thus constructed achieves a learning rate of the same order as that attained in absence of any sampling bias. Beyond these theoretical guarantees, numerical results providing strong empirical evidence of the relevance of the approach promoted in this article are displayed. |
| نوع الوثيقة: |
conference object |
| اللغة: |
English |
| Relation: |
hal-04431531; https://hal.inrae.fr/hal-04431531; WOS: 00683104600074 |
| الاتاحة: |
https://hal.inrae.fr/hal-04431531 |
| رقم الانضمام: |
edsbas.60F70EF1 |
| قاعدة البيانات: |
BASE |